Ask HN: Seriously, which AI coding thing is best?
3 comments
The reason everyone has a different answer is that everyone has unique needs and skills. You should be getting a handful of ideas from what other people do, but the key is try different options out yourself, making your own call on which setup actually works best for your personal working style.
Right now, I'm using Grok Code Fast 1 for coding tasks. I've found it to be many times faster and cheaper than Claude 4.5 or GPT 5 Codex while still being quite capable.
The "code-supernova-1-million" stealth model was more capable, but it is unfortunately no longer available through any of providers. Hopefully it gets publicly released soon. My guess is that it is an updated code model from XAI/Grok.
Having to switch models every few weeks isn't the most fun, but I can't turn down the productivity that comes with it. I wish I could just go back to Neovim, but I can't justify the lower productivity.
The "code-supernova-1-million" stealth model was more capable, but it is unfortunately no longer available through any of providers. Hopefully it gets publicly released soon. My guess is that it is an updated code model from XAI/Grok.
Having to switch models every few weeks isn't the most fun, but I can't turn down the productivity that comes with it. I wish I could just go back to Neovim, but I can't justify the lower productivity.
Here’s how I decide what to use:
ChatGPT – journaling, talking, planning. Codex – framework and middleware-layer coding. Claude Code – logic and application-level coding. Anthropic models via OpenRouter + Cline – when the task is error-prone, tedious, or needs high fidelity; lower error rate in my experience, though pricier. Cursor Agents – multi-file integration, boilerplate, and forking tasks.
Each fills a different slot in the workflow, so “best” depends on what kind of coding you’re doing.
ChatGPT – journaling, talking, planning. Codex – framework and middleware-layer coding. Claude Code – logic and application-level coding. Anthropic models via OpenRouter + Cline – when the task is error-prone, tedious, or needs high fidelity; lower error rate in my experience, though pricier. Cursor Agents – multi-file integration, boilerplate, and forking tasks.
Each fills a different slot in the workflow, so “best” depends on what kind of coding you’re doing.
Simply put, I want to know, today, which should I be using?